Viewpoints : The Chip Industry After U.S. Memories
A critical battle that will help determine U.S. industrial competitiveness against the Japanese is being fought over dynamic random access memory chips, or DRAMs. So far, the Americans appear to be losing.
DRAMs hold the key to computer memory. Without them, there are no advanced computers and there would likely be no digital televisions, stereos and other equipment that require large amounts of memory. But U.S. producers, who held 59% of this market at the beginning of the 1980s, now hold only 14%. Japanese market share, meanwhile, has risen to 78% from 39% in this period. If American companies can’t competitively make these chips, U.S. computer firms will be dependent on Japanese producers, further upsetting the balance of trade.
Last June, former International Business Machines Vice President Sanford Kane announced that he was heading an attempt to form a joint effort of American companies--called U.S. Memories--that would work together to reduce Japanese dominance of DRAMs. However, T. J. Rodgers, president of Cypress Semiconductor, a Silicon Valley company that makes another kind of chip, soon emerged as the venture’s chief critic, saying it was not the way to beat the Japanese.
Two weeks ago, Rodgers was triumphant. The U.S. Memories project was killed because of insufficient support and a temporary oversupply of DRAMs.
Free-lance journalist Sharon Bernstein interviewed Rodgers and Kane about the fall of U.S. Memories and the future of the U.S. chip industry in separate interviews, allowing each to respond to the other.
- Why did the semiconductor and computer industry--which had initially supported U.S. Memories--back out?
Kane: What went wrong is that today they (U.S. companies) are not in trouble. They didn’t want to make a firm commitment because they weren’t concerned about their future.
Rodgers: U.S. Memories died plain and simple because it was a bad business deal. A bad investment. The initial valuation of U.S. Memories was $500 million, so people were asked to pay $50 million for 10% of the stock. Well, for the same $500 million, they could buy two LSI Logics or one Cypress and have $100 million in change. A one-man corporation with no fabrication plants, no customers and no employees simply isn’t worth anything. Sandy Kane coming from IBM is not really capable of thinking about how he could start a company from scratch.
Kane: There’s no way to counter an argument from someone who knows none of the facts. What we were trying to do is respond to the industry need, and if you do this on a small scale, there’s no way you can afford it. There’s no way to do this efficiently on a small scale.
When I heard T. J. Rodgers’ comments about what he was going to do in Texas on a very small scale, I knew it was a joke. It costs a billion dollars to set up a plant capable of making 64-megabit chips.
- Should we be bracing for another DRAM shortage like we had a couple of years ago?
Kane: Is there likely to be another DRAM crisis? The only question is when.
The Japanese have already stated that they are cutting back on production of 1-megabit memory chips to stabilize prices. All they have to do is cut back a little bit more and prices go up even higher and you’ve got another shortage.
Another possibility is that as companies try to ramp down on production of 1-megabit memory chips and ramp up on 4-megabit memory chips--4-megabit chips are even more complex to make--it’s likely that there will be significant manufacturing problems. That could cause a shortage.
Currently, the only U.S. companies that are in the DRAM business are Texas Instruments, Micron and Motorola. Collectively those companies make less than 10% of the world’s supply of chips, 1-megabit and up. The only other company in the U.S. that makes memory chips is IBM, and they don’t sell them. They use them themselves.
In addition, with every generation of technology that goes by, the stakes go up. Fewer U.S. companies will make chips with higher amounts of memory because the size of the investment to make future generations of memory chips is going to get too expensive for all but the largest companies.
As you weed out the smaller guys, the ones most likely to remain are the large Japanese companies who are in this for the long haul.
Rodgers: That is a bogyman thrown up to frighten people. These guys go to Congress, scream (about the Japanese) and then get money and keep it for themselves. I think they’re concerned, but I think there’s a lot of self help in their drive to make America compete against the Japanese. And I think there are a lot of scare tactics thrown in with the concern.
- This past week, two American companies announced cooperative agreements with foreign companies to make or distribute DRAMs. IBM will work with the German electronics giant, Siemens, to develop a new-generation chip, and Intel has announced that it will sell DRAMs manufactured by Japan’s NMB Semiconductor Co. What is the significance of these developments?
Kane: Relative to U.S. Memories, they don’t mean anything. Neither of these things were done in response to the demise of U.S. Memories. One does not structure business arrangements on such a short-term basis.
Rodgers: They are significant because they prove that U.S. Memories was a bad idea, and not necessary. These are two open-market reactions to solve a problem that didn’t require any government intervention. Intel just gained economic control over millions of dollars worth of DRAMs. And IBM’s development deal with Siemens will make sure the Japanese don’t have domination over us. The open market has reacted appropriately to solve the problem.
It has meaning to me on a second plane too. Ten days ago, they were waving the American flag and telling their customers it’s a real good idea to give up $50 million to this consortium, and they had to be well along in their negotiations. Then they would have done an end run on their own consortium just days after it started, had it not failed.
- But doesn’t the cooperative nature of the ventures validate the U.S. Memories idea?
Kane: Yes. It is a superb indication of the fact that it is absolutely critical for companies to get together and start sharing their expertise and resources and lower the risks in development. As we look to the future of technology, the cost in technological skills as well as dollars of doing some of these advanced technologies is such that no company can assume they can do it on their own.
- Does this mean we’re out of the woods in terms of the DRAM problem?
Kane: Absolutely not. In the case of Intel, the scale of that venture is significantly smaller than U.S. Memories was going to be. And it’s not a technology exchange in any way. Intel is simply going to be a distributor for NMB, the Japanese company. And in spite of the fact that Intel will control the output of the facility in Japan and the fact that NMB is not a computer manufacturer and not a competitor of the U.S. companies, it’s still Japanese technology. It’s naive to presume that this company does not interact with other technology companies in Japan. U.S. customers will still be relying on foreign technology.
Rodgers: I think the DRAM market will work itself out globally. It’s a universally needed commodity. Market share is not an issue, as long as we’re not held hostage to the Japanese. And with all these other deals going on, we’re not going to be.
- Where do the U.S. computer and chip industries go from here?
Rodgers: Instead of having a semiconductor consortium, I’d rather have reduction in capital gains tax so we have better access to funds. The government could throw some money at education. And there should be absolute adherence to the anti-dumping law, the law that made it illegal for Japanese companies to flood the market and drive prices down.
Kane: There are a couple of things the government should be doing, none of them directly related to U.S. Memories.
First, I think it’s very important that they pass some antitrust legislation that would modify the existing laws to provide for a more conducive climate for joint manufacturing ventures. The way most people have been recommending changes would be to take the law that was passed five years ago on joint research and development ventures, and extend that to manufacturing capability. We need to make it so it’s not a violation of antitrust laws for companies to work together in manufacturing.
In addition, it is absolutely critical for the government to spend significant amounts of money in certain research areas, working with industry on development. The grants should be not at a product level, but at a basic research level.
The other area, which is less global but no less important, is that the government must take serious action to reduce the deficit. Until they do that, the cost of capital in the U.S. is never going to decrease.
